Detailed kinetic mechanism of gasphase reactions of volatiles released from biomass pyrolysis

Research paperDetailed kinetic mechanism of gas-phase reactions of volatiles released from biomass pyrolysis Paulo Eduardo Amaral Debiagi, Giancarlo Gentile, Matteo Pelucchi, Alessio Fra

Research paper

Detailed kinetic mechanism of gas-phase reactions of volatiles

released from biomass pyrolysis

Paulo Eduardo Amaral Debiagi, Giancarlo Gentile, Matteo Pelucchi, Alessio Frassoldati,

Alberto Cuoci, Tiziano Faravelli, Eliseo Ranzi*

Dipartimento di Chimica Materiali ed Ingegneria Chimica “G Natta”, Politecnico di Milano, Piazza Leonardo da Vinci 32, 20133 Milano, Italy

a r t i c l e i n f o

Article history:

Received 7 April 2016

Received in revised form

7 June 2016

Accepted 17 June 2016

Keywords:

Biomass pyrolysis

Biomass gasification

Fast pyrolysis

Lumped mechanism

Bio-oil

Tars

a b s t r a c t

Comprehensive chemical models to describe the behavior of biomass pyrolysis, gasification and com-bustion are crucial for the simulation and design of thermochemical processes of ligno-cellulosic ma-terials Despite this importance, reliable and predictive models are still not well known The original aspect of this work is to present a comprehensive and predictive model of pyrolysis, gasification, and combustion, starting from biomass characterization, through the description of released volatiles at the particle scale, until the effect of the secondary gas-phase reactions at the reactor scale All these aspects can play a relevant role in the biomass thermo-valorization processes Most of released species from biomass devolatilization are oxygenated hydrocarbons This study aims at identifying some reference rate parameters, based on analogy and thermochemistry rules, for the different reaction classes Once rate rules are defined, they allow an easy extension to analogous compounds In this way, the kinetic mechanism already developed for jet and diesel fuels is extended to the new tar species released by biomasses Despite unavoidable approximations when the interest is also at the reactor scale, this model

is the only one, to our knowledge, able to describe the whole process from biomass tofinal products, in a predictive and satisfactory way

© 2016 Elsevier Ltd All rights reserved

1 Introduction

Biomass thermal conversion processes produce heat, electricity,

and fuels Fast pyrolysis is a promising process for the production of

renewable bio-oils and chemicals[1] Bio-oil is a complex mixture

of anhydrous sugars, furan derivatives, oxygenated aromatics, and

low molecular weight products [2,3] Furthermore, the biomass

integrated gasification/combined cycle system is amongst the most

promising modern technologies, because of its higher energy ef

fi-ciency compared to direct combustion[4] One of the major issues

in biomass gasification is to deal with tar formed during the

pro-cess Similar to bio-oil, tar is also a complex mixture of condensable

hydrocarbons, which includes several oxygen-containing

hydro-carbons, along with phenolic and multiple ring aromatic

com-pounds Tar produced from gasification process mainly contains

compounds without oxygen (tertiary tar), as opposite to pyrolysis

[5] The composition of tar and bio-oil is a key factor in assessing

pyrolysis and gasification processes Indeed, it involves hundreds of organic compounds, and it depends on feedstock types, reactor temperature, residence time, and catalytic effects associated to ash The approximation of tar as a narrow range of components leads to inaccurate predictions of dew point temperature and gasification

efficiency Aiming at improving the understanding of pyrolytic behavior of biomass, Shen et al.[6]recently reviewed biomass fast pyrolysis discussing the yields of liquid and gas products, focusing

on the primary and secondary formation pathways of oxygenated compounds Despite the outstanding importance of developing a reliable reaction model for the design and optimization of biomass pyrolysis and gasification, many difficulties lie behind its

complexity and varieties of components found in biomass, together with involved reactions, are the main reasons of these difficulties Even if the major components of biomass are only three macro-molecules such as cellulose, hemicellulose and lignin, their relative compositions vary significantly[8], and only cellulose structure is well defined A further difficulty in developing a detailed kinetic mechanism for biomass pyrolysis is that reactions proceed simul-taneously in the condensed and gas phase, therefore their relative

* Corresponding author.

E-mail address: eliseo.ranzi@polimi.it (E Ranzi).

Contents lists available atScienceDirect Biomass and Bioenergy

j o u r n a l h o m e p a g e : h t t p : / / w w w e l s e v i e r co m / l o c a t e / b i o m b io e

http://dx.doi.org/10.1016/j.biombioe.2016.06.015

0961-9534/© 2016 Elsevier Ltd All rights reserved.

Biomass and Bioenergy 93 (2016) 60e71

role is not well defined.

The kinetic modeling of pyrolysis, gasification, and combustion

of biomasses is a very complex multi-component, multi-phase, and

multi-scale problem The strong interactions between chemical

kinetics, heat and mass transfer processes involved in the thermal

degradation of biomasses make the mathematical modeling dif

fi-cult These models require at least the following four features:

Characterization of the biomass in terms of reference

components;

Kinetics of devolatilization of reference species in the solid/

metaplastic phase;

Kinetics of the gas-solid gasification and combustion of residual

char;

Secondary gas-phase reactions of gas and tar species released

during devolatilization

While biomass characterization, together with the

correspond-ing multistep kinetic models of cellulose, hemicellulose, lignins,

and extractives has been recently reviewed [8], secondary gas

phase reactions of gas and tar released species are discussed in a

greater detail in this work Thus, Section2briefly summarizes the

characterization method, along with the multistep kinetic scheme

of biomass devolatilization Section3analyzes and discusses the

secondary reactions of gas and tar species released during biomass

pyrolysis Lastly, comparisons with experimental data highlight the

advantages and reliability, as well as the limitations, of the overall

biomasses

2 Biomass characterization and multistep kinetic model of

devolatilization

combustion processes Therefore, the characterization of pyrolysis

products has attracted great analytical interests[9] One and

or time-of-flight mass spectrometry techniques allow qualitative

and quantitative analysis of biofuels composition, thus describing a

large portion of bio-oil[10,11] Similarly, the application of tunable

synchrotron vacuum ultraviolet photoionization mass

spectrom-etry allows to improve the knowledge of this reaction system[12]

As biomass feedstocks are complex mixtures of several compounds,

it is very essential to characterize them in terms of specific number

of reference species While usual characterization methods are

limited to the study of cellulose, hemicellulose and lignin

compo-nents[13e15], detailed characterization was recently extended to

triglyceride and tannin species, representatives of extractives[8]

2.1 Biomass characterization

For this modeling work, the reference components of biomasses

are cellulose, hemicellulose, lignin, and extractives, which

consti-tute the largest portion of the biomass, often with ash Biomass

pyrolysis products are assumed as a linear combination of the

py-rolysis products of these reference compounds, neglecting their

possible interactions When direct information on biochemical

composition is not available, cellulose, hemicellulose, lignin, and

extractive content is derived through the ultimate biomass

composition in terms of H/C/O[8,15] As reference species, together

with cellulose and hemicellulose, three different types of lignins,

rich in carbon, hydrogen and oxygen, are considered[16] Finally,

triglycerides and condensed tannins are two lumped reference

species accounting for hydrophobic and hydrophilic extractives,

respectively.Table 1reports the seven reference species described

above The biomass composition in terms of the seven reference components is calculated from biomass elemental composition by solving the system of atomic mass balances for carbon, hydrogen and oxygen, together with a constraint that defines all fractions to

be positive The following system of linear equations expresses the three atomic balances:

CBM¼a$CRM1þb$CRM2þg$CRM3

HBM¼a$HRM1þb$HRM2þg$HRM3

OBM¼a$ORM1þb$ORM2þg$ORM3

where RM1, RM2and RM3are the reference mixtures anda,bandg

their corresponding fractions Reference mixtures are different combinations of the seven reference species RM-1 is a mixture of cellulose and hemicellulose, while RM-2 is a mixture of lignins with triglycerides andfinally RM-3 is again a mixture of lignins with some content of tannins Further details on this characterization method are reported in Debiagi et al.[8]

2.2 Multistep pyrolysis model Although the biomass composition and the thermal treatment conditions can significantly change the product distribution, a similar set of products is always obtained on a qualitative basis A peculiarity of this model is the detailed characterization of pyrol-ysis products, which not only includes water vapor and permanent gases (H2, CO, CO2, CH4, and C2H4), several alcohols, aldehydes, and carbonyl compounds, but also different sugars, phenolics and het-erocyclic species At high temperatures, several chemisorbed spe-cies contribute to char devolatilization progressively releasing CO2,

CO and H2 Each reference component decomposes independently through

a multistep, branched mechanism offirst-order reactions.Table S1

of the Supplemental Material reports the overall multicomponent and multistep kinetic mechanism of primary biomass pyrolysis This kinetic mechanism models the formation of intermediate solid and chemisorbed species, together with char, gas, and tar species Both cellulose and hemicellulose are polymeric sugar chains releasing, together with permanent gases, a wide number of hy-drocarbon and oxygenated species, including methanol, acetic acid, hydroxy-acetaldehyde, acetone, acetol, furfural, 5-hydroxymethl-furfural, levoglucosan and anydro-sugars[6] The multistep lignin

scheme of Faravelli et al.[16] These reactions are active in a broad temperature range and release phenolic components Phenol, ani-sole (metoxy-benzene), 2,6-dimethoxy-phenol, 4-(3-hydroxy-1-propenyl)phenol, and 3-(4-hydroxy-3,5-dimethoxy-phenyl) acryl-aldehyde are a few lumped and representative species of these compounds Phenol is also released by thefirst decomposition step

of tannins, while triglycerides (TGL) quickly decompose to a lum-ped species representative of free fatty acid

The rates and stoichiometries of these lumped reactions were originally derived from experimental findings [15] The kinetic model is continuously updated, based on new experimental data and comparisons across a wider range of experimental conditions Recently, experimental data showed the overshooting of temper-ature at the center of thick biomass particles, and allowed a better evaluation of the endothermic release of tars and the exothermic charring process[17]

While several detailed kinetic mechanisms are discussed in the

applied in this work is very simplified, aiming at an effective use not

P.E.A Debiagi et al / Biomass and Bioenergy 93 (2016) 60e71

only at the particle scale [17], but also at the reactor scale.

Computational time limitations are indeed very severe when

simulating a gasifier or a biomass combustor at the reactor scale

[21,22]

As a matter of a simple application example, the almond shell

analyzed by Caballero et al.[23]is considered.Table 2shows the

ultimate analysis of this biomass sample, together with the

corre-sponding detailed composition in terms of reference species.Fig 1

shows the comparison between model predicted and experimental

TG curve DTG curves of individual reference components are also

shown

Table 2 also reports model predictions in terms of detailed composition of released gas and tar species, together with the re-sidual char composition Experimental data onfinal products are not available, but model predictions in terms of detailed informa-tion on gas, tar, and residual char are reported in order to show how the model is able to describe pyrolysis products While the primary multi-step kinetic model of biomass pyrolysis was already vali-dated and discussed in recent papers[8,17], the next Section in-vestigates and analyzes the secondary gas phase reactions of the released species The products reported in Table 2 represent a sample of species whose secondary reactions are analyzed in the

Table 1

Reference species for biomass characterization (after Debiagi et al [8] ).

P.E.A Debiagi et al / Biomass and Bioenergy 93 (2016) 60e71

next Sections.

3 Kinetic scheme of secondary gas phase reactions

During biomass pyrolysis and gasification, primary volatile

products are often exposed to high temperatures where gas phase

decomposition or combustion reactions can play a significant role

These secondary reactions are a significant part of the overall and

comprehensive kinetic model of biomass pyrolysis and oxidation

However, reliable experimental thermochemical data for all these

species are not always available[24] Recently, the decomposition

products of lignins, such as monocyclic aromatic compounds with

substituent groups like hydroxy, methoxy, formyl, vinyl, and alkyl

received particular attention[25] Carstensen and Dean[7]gave a

very significant contribution in defining the secondary gas phase

reactions of volatiles released from biomass pyrolysis Fromfirst principle calculations and for a given reaction class, they system-atically studied kinetic laws on a series of small reactants and derived generic rate rules, to be extrapolated to all members of the same reaction class In facts, it is not feasible to perform high-level calculations for every reaction of this large kinetic model Namely, they analyzed the H-abstraction and water elimination reactions from alcohols, and the initial decomposition reactions of phenyl ethers

The gas-phase kinetic model discussed and applied in this paper was obtained by extending the POLIMI kinetic mechanism for the pyrolysis and oxidation reaction of hydrocarbon and oxygenated species[26]

Table 3 reports a list of relevant oxygenated species released from biomasses, whose primary decomposition reactions are here shortly discussed

Due to their modular structure, the extension of detailed kinetic mechanisms to the new species requires to include their primary propagation reactions Typically, the reaction classes to be included are initiation, H-abstraction and addition reactions, together with molecular and successive radical decompositions until the forma-tion of intermediate products already accounted for in the original kinetic scheme The complete and extended kinetic mechanism, as well as thermodynamic and transport properties, is reported as

Supplemental Material and is also available at http:// creckmodeling.chem.polimi.it/ Here, we simply and shortly sum-marize and revise the general rules applied in developing and extending the whole kinetic mechanism to the new species re-ported inTable 3

3.1 Chain initiation and H-abstraction reactions Since the pioneering developments of detailed oxidation

different reaction classes with well-defined rate rules is the first step towards the computer-aided generation of complex kinetic mechanisms of large hydrocarbon and oxygenated fuels Several reaction classes, together with physically consistent rate rules and kinetic parameters, were established, and these reaction classes and rate rules are continuously completed and revised based on new experimental and theoretical data[31,32]

Unimolecular initiation reactions activate the chain radical propagation process, breaking chemical bonds and forming two radicals:

Table 2

Almond Shells [23] Ultimate analysis, predicted characterization in terms of

reference species, and predicted composition of pyrolysis products from TGA of

Fig 1

Almond shell ultimate analysis (wt% dry and ash free)

50.9% 6.1% 43.0%

Composition in terms of Reference Species (wt%) (see Table 1)

CELL HCELL LIGH LIGO LIGC TGL TANN

44.6% 20.3% 7.7% 14.3% 6.3% 3.7% 3.2%

Yield and elemental composition of gases, tar, and solid residue (wt% of

initial sample) Gases 12.9 Tars 63.1 Solid residue 23.7

Yield of gas and tar species (wt% of initial sample)

Gases Tars

CO 4.80 H 2 O 9.6 C 2 H 4 O 2 2.8

CO 2 6.50 CH 2 O 2.5 Phenol 1.3

H 2 0.03 CH 3 OH 2.5 Anisole 0.9

CH 4 0.60 CH 3 CHO 0.6 Hydroxy-methyl-furfural 1.8

C 2 H 4 1.00 HCOOH 0.3 Xylosan 6.2

C 2 H 5 OH 0.3 Coumaryl alcohol 0.8 Acrolein 0.2 Levoglucosan 23.4 Glyoxal 0.7 Sinapyl Aldehyde 3.9 Propanal 1.5 Fatty acids 3.5

Fig 1 Pyrolysis of almond shell (2 K/min) Comparisons between experimental data (points) [23] and model predictions (lines) DTG curves of individual reference components are

P.E.A Debiagi et al / Biomass and Bioenergy 93 (2016) 60e71

R R’4R$ þ R’$

Rate constants strongly depend on the dissociation energy of

the bond involved in these reactions

H-abstraction reactions play a key role in determining the

sys-tem reactivity both in pyrolysis and oxidation conditions It is

generally accepted that the rate constant of the generic reaction:

R$ þ R’H4R’$ þ R  H

depends on the properties of the abstracting radical (R
) and the

type of hydrogen to be abstracted[33e35]

The type of H-atom mainly refers to the strength of the

corre-sponding C-H bond Then, prior to the discussion of the rate rules, it

is convenient to analyze the differences in the C-H and O-H bond

dissociation energies (BDEs) of different hydrocarbon and

oxygenated species Similarly, the BDEs of the C-C and C-O bonds

are useful to investigate the relative importance of the

unim-olecular initiation reactions

3.1.1 Bond dissociation energies (BDE) of C4hydrocarbon and

oxygenated species

The G4 computational method developed by Curtiss et al.[36]as

implemented in the Gaussian-09 suite[37]was used to calculate

C-H and C-C (as well as O-C-H and C-O) bond dissociation energies

According to the corresponding elementary decomposition

re-actions, the BDEs are determined at 298 K as the difference in the

G4-energy between the reactant and the formed decomposition

radicals.Table 4summarizes the BDEs for n-butane, iso-butane,

1-and 2-butene, together with n-butanol 1-and n-butanal as reference

oxygenated species The calculated BDEs are in good agreement

with the corresponding ones estimated with the MRACPF2 method

by Oyeyemi et al.[38] Further details on these BDE calculations are

reported by Pelucchi et al.[39]

3.1.2 H-abstraction reactions

The dominant reactive radicals in pyrolysis and oxidation

conditions are H, OH and CH3, which abstract H-atoms to form H2,

H2O and methane H-abstraction reactions from pure hydrocarbons have shown that individual rate constants are obtained, with good accuracy, by generic rate rules for individual rate expressions

[7,33,35,40] Due to the different C-H BDEs, the activation energy required by methyl and alkyl radicals for the abstraction of a sec-ondary H-atom in an alkane is ~2.5 kcal/mol lower with respect to the corresponding energy to abstract a primary H-atom Again, by comparing primary and tertiary H-atoms in iso-butane, the tertiary H-atom abstraction is favored by ~4 kcal/mol Whenever direct and more accurate kinetic parameters were not available, these energy corrections and simple rules have been successfully applied in the development of pyrolysis and oxidation mechanisms[26].Fig 2

shows the H-abstraction rates of primary, secondary, and tertiary H-atoms by H
,
OH and
CH3radicals Accordingly, with the large BDE differences of C-H bonds, H-abstractions on alkenes indicate that H-atoms in allyl and vinyl position become more and less reactive, respectively These rate values are also reported inFig 2, with respect to the secondary H-atoms These simple rules allow the automatic generation of the H-abstraction reactions for the whole class of alkanes and alkenes, properly accounting for the individual reactivity of the different H-atoms

The same approach is applied to alcohols and aldehydes, and more in general to the oxygenated tar species, where the presence

of oxygen atom largely influences the BDE of adjacent C-H bonds, as shown inTable 4 Previous kinetic studies on alcohol fuels[41,42]

clearly highlighted that the rate of H-abstraction from the hy-droxyl group (ROH) by a generic radical is even lower than the corresponding abstraction rate of a single primary H-atom This is fully consistent with the BDE of the RO-H bond in n-butanol, which

is higher than the BDE of the primary H atoms in alkanes Moreover, the nature of the alkyl R-group does not significantly affect these values Successive decomposition reactions of alkoxyl radicals are discussed by Curran[43]

H-abstractions from C-H bonds in thea-position to the hydroxyl group strongly depend on the type of the H-atom to be removed Rate parameters of the abstraction ofaH-atoms, such as those of

Table 3

Formation enthalpyDH f,298 [kcal/mol] and formation entropyDS f,298 [cal/mol/K] of major oxygenated species released from biomasses and involved in secondary gas phase reactions.

P.E.A Debiagi et al / Biomass and Bioenergy 93 (2016) 60e71

ethanol, n-propanol, 1-butanol, and iso-butanol, are ~1.5 faster than

the corresponding rate parameters of abstraction of secondary

H-atoms in alkanes (Fig 2) Similarly, the abstraction parameters ofa

H-atoms, such those of iso-propanol and 2-butanol, are ~1.5 times

the corresponding ones of tertiary H-atoms According to

Car-stensen and Dean[7], the influence of the OH group on the

reac-tivity of C-H bonds practically vanishes at theb-position From this

site on, the rate constants of the different H-atoms follow the reference values of H-atoms in alkanes

Due to the low BDE of the C-H bond in the carbonyl group of aldehydes, the removal of the acylic H-atom is highly favored Based on the analysis of these reactions on formaldehyde, acetal-dehyde, and heavier aldehydes[44], the following rate parameters for abstraction of the acylic H-atom by H, OH and CH3radicals are

Table 4

Bond dissociation energies (BDE) of C 4 hydrocarbon and oxygenated species C-H (black), C-C (red), and C-O & O-H (blu) bond dissociation energies (kcal/mol) calculated at G4 level (298 K) [39]

Fig 2 H-abstraction reactions Calculated rate constants (per H-atom) for simple primary, secondary, tertiary H-atoms (top) and for secondary H-atoms in alkyl, vinyl, and allyl positions (bottom).

P.E.A Debiagi et al / Biomass and Bioenergy 93 (2016) 60e71

kH¼ 2:5$1014expð6360=RTÞ
cm3

s=mol

kOH¼ 2:0$1013expð630=RTÞ
cm3

s=mol

kCH3¼ 7:0$1011expð7235=RTÞ
cm3

s=mol

The H-abstraction rates from C-H sites in a-position to the

carbonyl group again are enhanced by a factor 1.25 with respect to

the corresponding C-H sites in alkanes All these generic rate rules

are useful to create afirst reasonable set of rate parameters More

accurate evaluations could be of course required when rate and

sensitivity analyses identify sensitive reactions[7]

3.1.3 Unimolecular initiation reactions

C-C and C-H bond dissociation energies are very useful to

directly estimate the unimolecular initiation reactions These

re-actions are always the important chain initiation steps even if their

direct contributions to fuel consumption are only present at

rela-tively high temperatures[45] Assuming the rate constant for the

reverse recombination reactions, from the microscopic reversibility

principle it is practical to evaluate the rate constant of initiation

reactions If the activation energy of the recombination reactions is

zero, the BDEs directly become the activation energy of the

corre-sponding unimolecular dissociation reaction Of course, the favored

initiation reactions are the ones involving the lower activation

energies, i.e the lower BDEs The difference between the BDE of

1-butene and n-butane to form the resonant allyl radical and two

ethyl radicals, respectively, explains the different activation energy

of these two initiation reactions:

nC4H104$C2H5þ $C2H5 k¼ 1016expð  82000=RTÞ
s1

1C4H84$C3H5þ $CH3 k¼ 1016expð  73000=RTÞ
s1

Similarly, the activation energies of initiation reactions of

butanal[44]:

reflect the different dissociation energies of the corresponding

bonds

The initiation reactions, which cleavage C-H bonds, are mainly

important in the reverse direction as sinks of H atoms, because of

their very high BDEs

3.2 Carbohydrates and water elimination reactions The previous already mentioned kinetic studies of alcohol fuels highlighted the importance of the following molecular water elimination reactions:

These kinetic parameters, well confirmed by theoretical calcu-lations[46]as well by the recent review of Sarathy et al.[47], show that reference rate values for this reaction class are only slightly affected by the position of the OH group inside the carbon skeleton While the different alcohol has little impact on reference rate constant, large deviations are observed for substituted aldehydes, when water elimination reactions yield products with conjugated

dehydration reactions in glycerol pyrolysis[48] Thefirst dehydration reaction can either form prop-1-ene-1,3-diol or prop-1-ene-2,3-prop-1-ene-1,3-diol Keto-enol tautomerism transforms prop-1-ene-1,3-diol into 3-hydroxypropanal, which rapidly forms acrolein through a second dehydration reaction The aldehyde moiety strongly influences the reactivity by stabilizing the transi-tion state and the products[7] This stabilizing effect is accounted for with a reduction of the activation energy of more than 10 kcal/ mol

These reaction rates also dominate thefirst molecular dehy-dration with/without ring opening of levoglucosan and xylan components, as well as successive dehydration reactions Mayes

et al.[49] presented a very comprehensive study of elementary mechanisms of unimolecular glucose 1,2-dehydration reactions and conversion to the pyranose and furanose forms of levogluco-san, together with the glucose pyrolysis to form 5-hydroxy-methyl-furfural (C6H6O3: HMFU) The comparison of different dehydration reactions emphasizes the importance of adjacent functional groups and stereochemistry in determining reaction kinetics Successive

furfuryl-alcohol and (C5H6O2) furfural (C5H4O2)[6]

Along with the molecular dehydration, the unimolecular initi-ation and H-abstraction reactions are also considered for these species Intermediate radicals can then decompose forming the

hydroxyl-acetaldehyde, glyoxal, acetol, and other small-oxygenated compo-nents Retro-Diels Alder reactions constitute a parallel molecular

CH3CH2CH2CHO4$CH3þ $CH2CH2CHO k¼ 1016:3expð86000=RTÞ
s1

CH3CH2CH2CHO4$C2H5þ $CH2CHO k¼ 1016:0expð79000=RTÞ
s1

CH3CH2CH2CHO4$C3H7þ $CHO k¼ 1016:0expð80000=RTÞ
s1

1 C4H9OH41  C4H8þ H2O k¼ 1014expð67600=RTÞ
s1

2 C4H9OH42  C4H8þ H2O k¼ 1014expð66100=RTÞ
s1

2 C4H9OH41  C4H8þ H2O k¼ 1:5$1014expð67100=RTÞ
s1

iso C4H9OH4iso  C4H8þ H2O k¼ 5:0$1013expð65600=RTÞ
s1

tert C4H9OH4iso  C4H8þ H2O k¼ 4:5$1014expð65100=RTÞ
s1

P.E.A Debiagi et al / Biomass and Bioenergy 93 (2016) 60e71

path to form C2-C4oxygenated species[7] Aromatic and phenolic

species and successive reactions

Thefirst decomposition of lignin releases aromatic and phenolic

components Saggese et al [50,51] recently discussed the

inter-mediate and high temperature reactions of benzene with the

suc-cessive paths to form polycyclic aromatic hydrocarbons (PAH) and

soot While the reactions of PAH aromatics (involving naphthalene,

acenaphthylene, phenanthrene, pyrene, up to C20aromatics) are

considered in the gas-phase kinetic scheme reported in the

Supplemental Material, the soot kinetic model [50]contains the

successive reactions of PAH species to form carbon particles up

50e100 nm

Kinetic studies on phenol, cresol, and anisole chemistry show

the importance of CO elimination from unsubstituted and substituted phenoxy radicals The reference reaction rate raises from the following reaction:

C6H5O$/cyC5H5$ þ CO k¼ 5:0$1011expð43920=RTÞ
s1

Reactions of phenoxy-substituted species to form CO and cyclopentadienyl radicals were also revised by Carstensen and Dean [35] Successive reactions of cyclopentadienyl radicals are responsible for the formation of naphthalene and heavier PAHs

[52] While phenol and cresol were investigated for their interest in combustion systems, anisole (C6H5OCH3) was mainly studied as the simplest surrogate for primary tar from lignin pyrolysis[53,54] Chain initiation reactions of aromatic species containing a methoxy group (-OCH3) involve the breaking of the weak C-O bond

in the methoxy group (BDE ~63.2 kcal/mol)[55] Indicating with Ph the unsubstituted or substituted phenyl groups, the following reference reaction is assumed for the formation of phenoxy or phenoxy like radicals (Ph-O):

PhOCH3/PhO$þ$CH3 k¼ 3:0$1015expð63200=RTÞ
s1

different alkyl radical Together with the H-abstractions on side groups of aromatic species, the following rates are selected as reference values for the ipso-addition reactions on generic aromatic species:

Fig 4 Mole fraction profiles of major species during the stoichiometric oxidation of anisole at residence time 2 s and 106.7 kPa Symbols refer to experiments [54] and lines to

Fig 3 Water elimination reactions in glycerol pyrolysis.

H$ þ Toluene/Benzene þ $CH3 k¼ 1:2$1013expð5100=RTÞ
cm3

s=mol H$ þ Anisole/Phenol þ $CH3 k¼ 1:0$1013expð6000=RTÞ
cm3

s=mol H$ þ Phenol/Benzene þ $OH k¼ 1:2$1013expð6000=RTÞ
cm3

s=mol H$ þ Anisole/Benzene þ $OCH3 k¼ 1:0$1013expð8000=RTÞ
cm3

s=mol

$OH þ Toluene/Cresol þ H$ k¼ 1:1$1012expð11000=RTÞ
cm3

s=mol

$OH þ Toluene/Phenol þ $CH3 k¼ 4:4$1012expð6700=RTÞ
cm3

s=mol

$CH3þ Phenol/Cresol þ H$ k¼ 1:3$1012expð16200=RTÞ
cm3

s=mol

$CH3þ Phenol/Toluene þ OH k¼ 1:0$1012expð15000=RTÞ
cm3

s=mol P.E.A Debiagi et al / Biomass and Bioenergy 93 (2016) 60e71

All these reactions progressively convert the aromatic and

phenolic species and their rate values mainly derive from the

ki-netic studies of pyrolysis and oxidation of anisole and phenol

[51,53] More recently, anisole pyrolysis and oxidation were studied

by Nowakowska et al.[54]in a jet-stirred reactor under diluted

conditions at 673e1173 K, residence time 2 s, and 106.7 kPa.Fig 4

shows selected comparisons between these experimental data and

the predictions of the proposed model

Saggese et al [51] observed a similar satisfactory agreement

when discussing and comparing model predictions with

reactors

4 Comparisons with experimental data

Since the previously referred papers [15,17], several

experi-mental data at particle and reactor scale were used to develop and

validate the overall biomass pyrolysis model, including the

sec-ondary gas-phase reactions Fast and slow pyrolysis[56], severe

gasification in drop tube[57], combustion in a traveling grate[58]

are only a few examples of the analyzed experimental conditions

Here the comparisons are limited to recent experimental data,

which give more attention to the secondary gas phase reactions of

released species Norinaga et al.[59]and Yang et al.[60]studied the

kinetics of secondary vapor-phase decomposition of volatiles

generated from the fast pyrolysis of cellulose and lignin in a

two-stage tubular reactor, while minimizing volatile-char interactions

Avicel cellulose with particle sizes ranging from 74 to 105mm was

used by Norinaga, while the lignin samples used by Yang had

particle sizes in the range of 75e150mm These particle sizes allow

to neglect mass and heat transfer limitations Norinaga et al.[59]

provided experimental data of tar and gas compositions during

secondary pyrolysis of cellulose volatiles, at temperatures of 700

and 800C The most abundant product is always CO, along with

major species such as H2O, CH4, and H2 These data are useful not

only to verify the primary released species, but mainly to analyze

the effect of the secondary gas phase reactions studied with

resi-dence times up to 6 sFig 5shows a satisfactory comparison

be-tween experimental data and model predictions in terms of time

evolution of major pyrolysis products

It should be observed that the high temperature profile of

lev-oglucosan is practically zero, because of its very high reactivity, as

also experimentally observed On the contrary, the predicted acetone decomposition at low temperatures is lower with respect

to the wide scattered experimental measurements, which range between 0.0015 and 0.0044 The reliability of the kinetic model of acetone decomposition was already proved in comparison with pyrolysis data of acetone-butanol, and ethanol (ABE) mixture[61] The kinetic model is able to predict with reasonable accuracy the formation of aromatic species This is a mechanistic con firma-tion that aromatic hydrocarbons are mainly the result of successive condensation reactions, in agreement with the experimental and theoretical analysis of Norinaga et al [59] This feature will be emphasized in the successive application example, where higher temperatures and severity conditions further promote polycyclic-aromatic hydrocarbons (PAH) and soot formation

Very recently, Yang et al.[60]investigated the vapor-phase re-actions of nascent volatiles derived from the fast pyrolysis of lignin The two-stage tubular reactor was used for evaluating the sec-ondary gas phase reactions of the released species at temperatures from 500C to 900C, at 241 kPa The minimum residence time of volatiles before the detector was 0.1 s, while it was modified inside the secondary pyrolysis reactor up to 3.6 s The lignin samples were prepared by enzymatic hydrolysis (EHL) of empty fruit bunches, with the elemental C/H/O composition of 63.5/5.93/30.57 (on a dry basis) Fast pyrolysis was realized in thefirst isothermal reactor, and well-resolved chromatograms were obtained in the entire tem-perature range The characterization of the primary volatile prod-ucts released by lignin includes a large amount of heavy undetectable phenolic species (>30% at 773 K), and this is a clear

model predictions In simulating their experimental data, Yang

et al.[60]observed that it was very helpful to add a global and empirical reaction to account for the contribution of tar and un-detectable products in the gas-phase reactions Their predictions without the global reaction largely under-estimated the yields of several major products Therefore, they estimated the rate constant and stoichiometric coefficients of the global reaction by fitting, in a different way at the different temperatures, the experimental ob-servations and the predicted results

Table 5reports the primary volatile products released from fast pyrolysis of lignin as experimentally measured after 0.1 s[60] Very heavy phenolic species are not considered in the gas-phase kinetic scheme, because of the lumping approach For this reason, the experimental data are here corrected by assuming the undetected heavy species as equally distributed between the lumped phenolic

Fig 5 Effect of secondary gas-phase reactions on volatile species released from cellulose pyrolysis at 700C and 800C Comparison between experimental data (symbols) [59] and model predictions (lines).

P.E.A Debiagi et al / Biomass and Bioenergy 93 (2016) 60e71

species and the solid char residue.

Predicted data, as obtained directly from lignin pyrolysis, and after 0.1 s at the pyrolysis temperature, are also reported Com-parisons between experimental data and model predictions indi-cate an overall reasonable agreement The effect of secondary gas phase reactions with 0.1 s residence time is well evident at high

oxygenated species to form CO and H2, as well as a relevant for-mation of aromatic species, up to heavy PAHs

Fig 6shows the good agreement between experimental data and model predictions in terms of carbon, oxygen and hydrogen content in the char residue, as a function of the pyrolysis temper-ature As expected, the oxygen and hydrogen content progressively decreases with increasing temperature

Fig 7compares experimental and model predictions of volatile species released from lignin pyrolysis at different temperatures with residence time of 3.6 s in the second tubular reactor[60] Two different sets of model predictions are reported inFig 7 The former set refers and uses the experimental information of product dis-tribution from primary lignin pyrolysis reported inTable 5 Phenolic species, together with the undetected ones, were distributed ac-cording to the predicted primary pyrolysis products of the EHL

Table 5

Primary volatile products released from fast pyrolysis of lignin at 0.1 s Comparisons between experimental data [60] and model predictions.

Gas phase (0.1 s) Primary pyrolysis þ0.1 s of secondary

reactions

Primary pyrolysis

PRODUCTS (wt% of initial)

a With 50% of undetectable products added.

Fig 6 Carbon, hydrogen, and oxygen content in the residual char as a function of the

pyrolysis temperature Comparisons between experimental data (symbols) [60] and

model predictions (lines).

Fig 7 Volatile species released from lignin pyrolysis at different temperatures with residence time of 3.6 s in the second tubular reactor Comparison between experimental data

P.E.A Debiagi et al / Biomass and Bioenergy 93 (2016) 60e71